A sixteen-step sequence that holds attention for eight minutes is not a melody. It is a machine with four moving parts, and three of them are not the notes. Here is what each part does, and how to set it up on the KeyStep Pro.
Almost every other sequencer style is defined by where the rests go. Berlin School has none. Sixteen steps, sixteen notes, no gaps — which sounds like it should be unbearable and instead is the whole point. Once the ear accepts that nothing is going to stop, it stops listening for events and starts listening for change. That is the state the music is trying to get you into.
Tangerine Dream got there on Phaedra in 1974 with a Moog modular and a 960 sequencer, which had eight steps and no memory. Klaus Schulze pushed the same idea to twenty-five minute pieces on Timewind. Manuel Göttsching built E2-E4 from one sequence and almost an hour of patience.
Nothing arrives and nothing leaves. If you find yourself waiting for the pattern to do something, the sequence is right and the patch is not — the movement is supposed to come from the filter and the delay, not the notes.
The most common mistake is writing a good melody. Berlin sequences are built from a tiny pitch set — commonly three to five notes — repeated obsessively. Phaedra's line is essentially four pitches. A sequence that wanders through the whole scale stops being a cell and starts being a tune, and a tune demands you follow it somewhere.
Use the root, the minor third, the fourth and the fifth. Leave out the leading tone entirely: it points somewhere, and Berlin School is not going anywhere. The generator on this site enforces this — the Berlin style draws from a fixed pool of four scale degrees rather than the full scale.
This is the mechanism most people miss, and it is the one that does the work. Accent every fourth step and you get a bar — square, obvious, and boring by the third repeat. Accent every third step against a four-step bar and the emphasis takes twelve steps to come back round. The same sixteen notes then sound different on every pass, because the weight has moved.
On the KeyStep Pro this lives in velocity. The sequencer records velocity per step, so the accent pattern is baked into the sequence rather than played live — set your synth's velocity → filter cutoff modulation to about 40% and the accents open the filter as they pass. That is the entire "moving" quality of a Berlin patch.
If velocity is not routed to cutoff, the accents are just louder notes and the sequence stays flat no matter how well you programmed it. Check this before blaming the pattern.
Here is the thing that separates a Berlin sequence from sixteen repeated notes: the delay is not an effect, it is a second sequencer. Set a dotted-eighth delay against a sixteenth-note sequence and the repeats land in the gaps between your notes, producing a cross-rhythm that you did not program and cannot play. Three against four, generated by the echo.
That is the sound. Not the notes — the interference pattern between the notes and their own echoes.
Ping-pong if you have it. Darken each repeat or the top end turns to mush after four passes.
The pre-delay matters more than the decay. Without it the reverb swallows the attack and the sequence loses its edges.
A dotted eighth is three sixteenths. Against a sixteenth sequence, each echo lands three steps later — so the repeat of step 1 arrives on step 4, the repeat of step 4 on step 7, and the pattern of echoes walks through the bar exactly like the accent cycle does. The two rotations are the same trick applied twice.
The KeyStep Pro is unusually well suited to this, because it has four independent tracks with their own time divisions and a per-step velocity and gate that survive being recorded. Here is a working four-track layout.
| Track | Setting | Why |
|---|---|---|
| Track 1 — sequence | 1/16, 16 steps | The main cell. Route to your brightest voice; this is the one the delay hits. |
| Track 2 — counter | 1/16, 15 steps | The same idea one step shorter. It drifts against track 1 and realigns every 240 steps — Schulze's whole method in one setting. |
| Track 3 — bass | 1/4, 4 steps | Root and fifth only. Something has to stay still or the drift has nothing to drift against. |
| Track 4 — accent/perc | 1/8, 12 steps | Optional. 12 against 16 gives a slow three-against-four that never lands twice in the same place. |
| Swing | 50% (off) | Berlin is dead straight. Swing is the wrong idiom entirely — it makes it sound like house. |
| Gate | short, 25–40% | Short gates let the delay repeats speak in the gaps. Long gates smear over them and you lose the cross-rhythm. |
| Randomness | 5–12% | A trace only. This is the KeyStep standing in for analogue drift — enough that no two passes are identical, not enough to hear as an event. |
| Time Shift | 2–6% | Same reasoning. Real Moog sequencers were never perfectly in time and it is part of why they sound alive. |
| Lst Step | live control | The performance move. Shortening the sequence from 16 to 12 to 7 mid-piece re-phrases everything without touching a note. |
16 against 15 is polymeter — one pulse, different bar lengths, parts sliding out of phase over minutes. It is not polyrhythm, which would be conflicting pulses in the same span. Berlin School is almost always the first one, and the drift is slow on purpose.
Nothing above changes over time, and a Berlin piece is entirely about change over time. The sequence is the starting position, not the piece. In order, over eight to twenty minutes: